US20030223000A1

US20030223000A1 - Method for processing signals representing colour pictures

Info

Publication number: US20030223000A1
Application number: US10/445,480
Authority: US
Inventors: Jan Rooy
Original assignee: Individual
Current assignee: GVBB Holdings SARL
Priority date: 2002-05-28
Filing date: 2003-05-27
Publication date: 2003-12-04
Also published as: JP4502306B2; DE60300439D1; KR100988223B1; DE60300439T2; KR20030091674A; EP1367836B1; MY130516A; US7170554B2; TWI222314B; TW200307451A; JP2004005685A; CN1462151A; CN1320827C; EP1367839A1; EP1367836A1

Abstract

A method for processing at least a part of a colour picture comprises at least one pixel defined by a plurality of colour values. Each colour value is variable between black and a colour clip value.

The method has the following steps:

determining for each colour value the quantity of the colour value above the common clip value,

computing a linear combination of said quantities,

for at least one of the colour values, outputting a processed colour value based on the linear combination.

Description

FIELD OF THE INVENTION

The invention relates to a method for processing signals representing colour pictures.

BACKGROUND OF THE INVENTION

It is current practice for cameras to use 3 CCD arrays to capture a colour picture. Each array senses the picture through a given colour filter. The 3 colours which are commonly used for this purpose are red (R), green (G), and blue (B).

Each CCD array outputs a video signal carrying information about the light received by each of its pixels. The 3 video signals are then generally digitised and put together as a digital stream.

The light received by such a camera from a given direction is thus defined by 3 colour values, for instance between 0 and 4,095 if coded on 12 bits. These data are then further processed to get a usable video stream. The further processing steps can take place either in the camera itself or in a studio.

One of these processing steps is applying a white balance. The goal of the white balance is to determine the weight of the various colours to get a picture with a desired look, generally as natural as possible. It consists in applying a gain specific to the colour to each colour value.

As the range within which colour values out of CCD arrays vary is common to the 3 colours whereas the gains for white balance are generally different (depending on the colour), the range of colour values after white balance is specific to each colour.

Differently said, the maximum value (corresponding to saturation of the CCD array and generally called clipping level) after white balance is generally different for the various colours.

Due to this difference, which can be important (e.g. a gain of 14 for red, 5 for green and 8 for blue), a big part of the dynamic range (after white balance) from the colours with highest clipping level remains unused.

SUMMARY OF THE INVENTION

The inventor proposes to make a better use of the dynamic range available in colour values to improve the dynamic range of the camera in the highlights.

In this goal, the invention proposes a method for processing at least a part of a colour picture comprising at least one pixel defined by a plurality of colour values, each colour value being variable between black and a colour clip value, with the following steps:

determining the lowest of the colour clip values as common clip value,

computing a linear combination of said quantities,

for at least one of the colour values, outputting as processed colour value:

the colour value plus said linear combination if said colour value is below the common clip value;

the common clip value plus said linear combination if said colour value is above the common clip value.

According to a possible solution, the method applies to each colour value.

According to another possible solution, the method applies to the green value only, whereas the red value and blue value remain untouched.

According to a preferred embodiment, said linear combination is a mean value of said quantities.

Further features of the invention will appear in the following description of exemplary embodiments.

DETAILED DESCRIPTION OF PREFERED EMBODIMENTS

The proposed method generally applies to colour values (R[0021] _in, G_in, B_in)_i,jdefining pixels (i, j) after a white balance.
If values output by CCD arrays for pixel (i, j) and digitised are called (R[0022] _CCD, G_CCD, B_CCD)_i,j, the effect of the white balance for each pixel (i, j) can be written as:
(R _in)_i,j =R _gain·(R _CCD)_i,j
(G _in)_i,j =G _gain·(G _CCD)_i,j
(B _in)_i,j =B _gain·(B _CCD)_i,j.
For instance, R[0023] _gain=14, G_gain=5 and B_gain=8 can be used.
Each value (R[0024] _CCD)_i,j, (G_CCD)_i,jand (B_CCD)_i,jis for instance coded on 12 bits and can consequently take values between 0 and 4095. These values can be directly output from the CCD arrays, but they can also be retrieved from a recording of the raw data output from the CCD arrays. In this last case, a transfer curve can be applied at recording (to get only values of 10 bits on the recording medium); the inverse transfer curve is then applied at retrieval.
The limiting levels for each colour, called hereafter colour clip value, are thus: [0025]
R _max =R _gain·4095
G _max =G _gain·4095
B _max =B _gain·4095.
It is clear from above that R[0026] _max, G_maxand B_maxare generally different from one another.
The minimum of theses clip levels or colour clip values will be called herebelow common clip value and is thus defined as: [0027]
CLIP=min(R _max ,G _max ,B _max).
Above this common clip value CLIP, the colour information becomes unreliable. [0028]
The information above the common clip value CLIP is used as colourless enhancement signal as described below. [0029]
In a first embodiment of the method, the processed colour values (R[0030] _enhanced)_i,j, (G_enhanced)_i,jand (B_enhanced)_i,jfor each pixel (i, j) are computed as follows from the colour values output by the white balance step.
If (R _in)_i,j<CLIP then (R _clipped)_i,j=(R _in)_i,jelse (R _clipped)_i,j=CLIP.
If (G _in)_i,j<CLIP then (G _clipped)_i,j=(G _in)_i,jelse (G _clipped)_i,j=CLIP.
If (B _in)_i,j<CLIP then (B _clipped)_i,j=(B _in)_i,jelse (B _clipped)_i,j=CLIP.
The values (R[0031] _clipped)_i,j, (G_clipped)_i,jand (B_clipped)_i,jcan be called the clipped RGB data.
The colourless enhancement signal for each pixel is a linear combination: [0032]
(ENH_nocolour)_i,j=0.5·{(R _{in i,j} −R _{clipped i,j})+(G _{in i,j} −G _{clipped i,j})+(B _{in i,j} −B _{clipped i,j})}.
For each pixel, the colourless enhancement signal (ENH_nocolour)[0033] _i,jis added to the clipped RGB signal to obtain the processed (and enhanced) colours values
(R _enhanced)_i,j=(R _clipped)_i,j+(ENH_nocolour)_i,j
(G _enhanced)_i,j=(G _clipped)_i,j+(ENH_nocolour)_i,j
(B _enhanced)_i,j=(B _clipped)_i,j+(ENH_nocolour)_i,j.
Clearly, the colourless enhancement signal (ENH_nocolour)[0034] _i,jallows the colours with low clipping level (green and blue in the above example) to take higher values than the colour clip level, which results in an increase in the dynamic range.
The processed colour values can then be used (R[0035] _enhanced)_i,j, (G_enhanced)_i,jand (B_enhanced)_i,jinstead of the colour values (R_in, G_in, B_in)_i,jfor further processing.
The second embodiment of the method described below is particularly adapted to pictures representing blue sky with clouds. The difference between blue sky and white clouds is basically a modulation in the red signal. To maintain the blue sky we only compensate for the lack of green signal caused by clipping in the green channel. Using the colourless enhancement signal proposed above, the processed colour values in this embodiment are: [0036]
(R _enhanced)_i,j=(R _in)_i,j
(G _enhanced)_i,j=(G _clipped)_i,j+(ENH_nocolour)_i,j
(B _enhanced)_i,j=(B _in)_i,j.
The green signal only is corrected; red and blue signals remain untouched. [0037]
As previously, by increasing the dynamic range of the green signal after processing, the dynamic range of the camera as a whole is improved. [0038]

Claims

What is claimed is:

1. A method for processing at least a part of a colour picture comprising at least one pixel defined by a plurality of colour values,

each colour value being variable between black and a colour clip value, with the following steps:

determining the lowest of the colour clip values as common clip value,

computing a linear combination of said quantities,

for at least one of the colour values, outputting as processed colour value:

2. A method according to claim 1, with the steps of:

outputting as processed colour value for each colour value:

3. A method according to claim 1, wherein said colour values are a red value, a green value and blue value, with the steps of:

outputting as processed green value:

the green value plus said linear combination if said green value is below the common clip value;

the common clip value plus said linear combination if said green value is above the common clip value;

outputting the red value as processed red value;

outputting the blue value as processed blue value.

4. A method according to claim 1, wherein said linear combination is a mean value of said quantities.

5. A method according to claim 1, wherein said colour values are generated in a white balance step.

6. A method according to claim 1, with the step of:

digitising signals from CCD arrays and applying a white balance thereto to generate the colour values.

7. A method according to claim 1, wherein the colour values are retrieved from a previous recording.

8. A method according to claim 7, wherein an inverse transfer curve is applied to generate the colour values.